After the Famine: Emigration from

Ireland, 1850-1913
TIMOTHY J. HATTON AND JEFFREY G. WILLIAMSON

This article examines the determinants of emigration from post-Famine Ireland.

As Irish real wages rose relative to those in destination countries, the emigration
rate fell. We argue, from time series analysis, that much of the secular fall in the
rate can be explained by that narrowing of the wage gap. County-level, cross-
sectional analysis of emigration rates indicates that poverty and low wages, large
family size, and limited opportunities to acquire smallholdings all contributed to
high rates of emigration. Changes in those variables over time reflect the rise in
living standards, consistent with time series evidence.

B etween 1850 and 1913 more than 4.5 million men and women left
Ireland for a new life overseas. Even after the effects of the Great
Famine of the 1840s had largely disappeared the emigrant flood contin-
ued; the number leaving ultimately amounted to about five times the
number who died in the Famine. As a proportion of the population, the
rate of emigration from Ireland was more than double that of any other
European country, with as many as 13 persons per 1,000 emigrating on
average each year. Largely as a consequence of this mass emigration,
the Irish population fell from 6.5 to 4.4 million between 1851 and 1911.
In this article we explore the causes of that emigration, first examining
trends in the emigration rate and asking why it declined over time, then
explaining fluctuations in annual time series for both total emigration
and emigration to the main destination country, the United States. We
then examine the composition of the emigrants by age, sex, and region
of origin before analyzing emigration rates at the county level for four
census years, in an attempt to learn more about the domestic factors
that drove so many Irish abroad.

TRENDS IN IRISH EMIGRATION

The principal source of Irish emigration data is the emigration reports

of the Registrar General for Ireland. These statistics were collected and
recorded by police at the ports of embarkation from 1851 onward. The
The Journal of Economic History, Vol. 53, No. 3 (Sept. 1993). © The Economic History
Association. All rights reserved. ISSN 0022-0507.
Timothy J. Hatton is Reader in Economics at the University of Essex, Wivenhoe Park,
Colchester, CO4 3SQ England. Jeffrey G. Williamson is Laird Bell Professor of Economics,
Harvard University, Cambridge, MA 02138.
We are grateful for support from National Science Foundation Grant No. SES-9021951. We have
benefited from useful comments by Roy Bailey, George Boyer, Cormac 6 Grdda, and two
anonymous referees. We also thank Chris Cook, Theo Charitidis, and Boris Simkovich for research
assistance and Fabio Schiantarelli and Lisa Lynch for their hospitality.

575
576 Hatton and Williamson

35 T

30

20

10

H-H 111111 n 11111 11 1111111111111 1111111 111111

1B50 1860 1870 1880 1800 1900 1910
Year
FIGURE 1

IRISH EMIGRATION, 1852-1913

Source: Ferenczi and Willcox, International Migrations, p. 730.

records are for gross emigration only, but return migration to Ireland
was relatively rare: return migrants from the United States comprised
only about 6 percent of the outflow.' The data are plotted in Figure 1 as
a ratio per 1,000 of the total Irish population. The figure documents a
sharp fall in emigration rates following the Great Famine. The rate fell
from about 15 per 1,000 in the 1860s to about 8 per 1,000 in the years
immediately before World War I—rates that were nonetheless high by
the standards of the time.
A number of authorities have pointed out that these data substantially
underestimate the true level of outmigration, chiefly because of incom-
plete recording of the movement from Ireland across the Irish Sea to
Great Britain. Cormac 6 Gr&da, for example, applied survival rates to
estimates of the flow of migrants to Great Britain and found a shortfall
of around half a million between 1851 and 1911. The true figure for
migrants moving across the Irish Sea was probably double what the
1
Fitzpatrick, Irish Emigration, 1801-1921, p. 7.
 Irish Emigration, 1850—1913 577

TABLE 1
EMIGRATION FROM IRELAND TO VARIOUS COUNTRIES, 1876-1913
Irish Emigration to
Immigration to
Years United States United States Canada Australia New Zealand Great Britain
1876-1880 28,356 27,924 1,300 3,408 2 ,159 (33,901)
1881-1885 69,080 62,736 5,497 2,234 753 (18,480)
1886-1890 56,016 59,965 2,460 3,319 167 (9,832)
1891-1895 45,440 46,494 842 927 109 (4,306)
1896-1900 32,243 33,006 475 779 59 (6,878)
1901-1905 36,819 30,768 1,447 418 91 (8,634)
1906-1910 30,996 24,125 3,538 542 165 (6,224)
1911-1913 27,622 21,411 5,980 841 192 (2,012)
Sources: Ferenczi and Willcox, International Migrations, p. 731; and 6 GrSda, "A Note," p. 44.

statistics suggest, and perhaps even more.2 Hence, emigration to Britain

was probably a fifth to a quarter of the total rather than an eighth.
Emigration statistics are broken down by country of intended desti-
nation from 1876 on. These are given as five-year averages in Table 1.
Following 6 Grada, we have doubled the recorded numbers of emi-
grants across the Irish Sea.3 Yet even with this adjustment, the
outstanding feature of the table is the overwhelming dominance of the
flow to the United States. For purposes of comparison, Table 1 also
shows five-year averages of the number of Irish immigrants reported as
arriving in the United States for the same periods.4 Despite differences
in definition and source of the statistics, the orders of magnitude are
similar between the two series (though significant discrepancies emerge
after the turn of the century). Examination of year-to-year fluctuations
also indicates a close association between the two series.5
Although there are some biases in the statistics, these are unlikely to
undermine the conclusion that the emigration rate fell significantly over
2
Because the country of destination of emigrants was not distinguished in the emigration
statistics until 1876, 6 Gr£da ("A Note") estimated the recorded flow to Britain over the whole
period from 1852 to 1911 by subtracting an estimate of total extra-European emigration (based on
immigration data) from the recorded total emigration. This gave a total flow of slightly over half a
million to Great Britain. Assuming that this flow followed the same profile over time as did total
emigration, and depleting each cohort, he calculated that more than 1 million Irish must have
emigrated to Britain in order for the stock of Irish-born living in Britain to match the number
recorded at the 1911 census. Given the assumptions, this is probably a lower-bound estimate.
3
Ibid.
4
Statistics of the numbers arriving in the United States by country of origin are available from
1820 on. Until 1868 they relate to "alien passengers" and thereafter to "alien immigrants." Before
1872 a considerable number of those arriving from the United Kingdom were not attributed
specifically to England, Scotland, Wales, or Ireland. Hence, before that date the totals for Ireland
probably underestimate the true numbers of Irish immigrants.
3
The correlation coefficient between the two series is .945, and a regression of the immigration
series on the emigration series gave the following result ((-statistics are in parentheses):
EM(US)= -406.58 + 0.93 1M(US) R2 = 0.S9
(0.16) (17.34)
As the series for emigration is for calendar years and the series for immigration for fiscal years, it
is likely that a more exact alignment of the data would give an even closer fit.
578 Hatton and Williamson

time. The statistics suggest a decline in the numbers emigrating to

Britain up to 1890. If there were a relatively constant rate of underenu-
meration of emigrants to Britain, then the true emigration rate would
have fallen even faster than Figure 1 suggests. Any improvement over
time in the accuracy of the data would also indicate a steeper decline in
the emigration rate. Even if emigration to Great Britain were discounted
altogether, there would still be a decline in the emigration rate. It would
also be true for the United States alone, though the decline would be a
little less dramatic.
What accounts for the observed fall in Irish emigration? One sugges-
tion is that many Irishmen feared the recurrence of famine, but that this
fear diminished as memories of the Great Famine faded. Even before
the Famine, the Irish emigration rate was probably the highest in
Europe, at about 7 per 1,000 per annum between 1821 and 1841.6 Thus,
the level of emigration immediately before World War I can be viewed
as a return to the pre-Famine rate. Some have suggested that the
declining rate was linked to the changing status of Catholics under
British rule. According to Robert Kennedy, land and religious reforms
were accompanied by growing economic opportunity for Catholics, not
to mention the rise of Irish nationalism. Thus, "the achievement of Irish
independence in 1921 climaxed a set of changes which reduced the
reasons for Irish Catholic emigration from Ireland." 7 However, this
view is no longer widely held and has received little support in more
recent studies of Irish emigration, such as those of David Fitzpatrick
and Kerby Miller.8 According to the latter, "the vast majority of
Catholic emigrants left home for essentially mundane reasons similar or
identical to those which produced mass migration from other European
countries." 9
Another possibility is that the share of the population in the mobile
age range decreased, but this seems unlikely to account for much of the
decline. Though the proportion of the population in the mobile age range
(15 to 34 years) fell slightly—from 35.3 percent of the population in 1861
to 32.7 percent in 1911—this fact can account for at most a fall of 1 per
1,000 in the aggregate emigration rate. 10 Other demographic changes,
however, might have influenced the emigration rate, and we shall
6
Mokyr, Why Ireland Starved, p. 35.
7
Kennedy, The Irish: Emigration, Marriage and Fertility, p. 39.
8
Fitzpatrick, "Irish Emigration" and Irish Emigration, 1801-1921; and Miller, Emigrants and
Exiles and "Emigration as Exile."
9
Miller, "Emigration as Exile," p. 339.
10
If we take a starting point of 15 per 1,000 of the population in 1861 and divide this instead by
the population between 15 and 34, we get a rate of 42.5 emigrants per 1,000 between those ages.
Applying that in 1911 gives us an emigration rate of 13.9 per 1,000 of the total population. Hence,
the change in the age structure of the population cannot have accounted for more than a 1.1 per
1,000 drop in the total emigration rate.
 Irish Emigration, 1850-1913 579

discuss them further. Other forces, such as the declining cost of

emigration, may have had modest effects in the other direction. The cost
of a passage to North America fell in the early years up to the 1870s and,
relative to Irish wages, declined by more than half between 1850 and
1913.
A more important but sometimes neglected explanation for the
decline in Irish emigration is that Irish economic conditions improved
dramatically over the period relative to the countries of prospective
emigration. As 6 Gr&da has reminded us, "It must not be forgotten that
no matter how poor people were, living standards were rising so that,
other things being equal, there was less incentive to move. . . .Thus the
period between the famine and the Great War may be characterised as
one of steady improvement in bread and butter terms. And indeed the
recorded emigration rate to America decreased during it." 11 Most of the
Irish who emigrated were employed as unskilled workers. It makes
sense, therefore, to compare the urban unskilled real wage in the
destination countries with that prevailing in Ireland. Real wage rates for
urban unskilled laborers have recently been constructed for Ireland and
all those countries to which the Irish emigrated, and for all the years of
interest to us in this study.12 The ratios of American and British to Irish
real wages are graphed in Figure 2, those of Australian and Canadian to
Irish in Figure 3. Figure 2 shows that both the American and British
wage ratios relative to the Irish fell sharply over the period. The ratio of
American to Irish wage rates fell from 2.85 in 1850/54 to 1.67 in 1895/99
before rising slightly to 1.85 in 1909/13. The British/Irish ratio also
declined, from 1.71 in 1850/54 to 1.16 in 1909/13. The Australian/Irish
wage ratio fell from 2.51 in 1870/74 to 1.44 in 1909/13. But the
Canadian/Irish wage ratio increased slightly over the same period, from
2.18 to 2.28.
If the wage ratio is the appropriate measure of the relative attractive-
ness of emigration compared with staying at home, then these trends
should help explain why the Irish emigration rate fell over time.
Compared with the two major emigration destinations (America and
Britain), all of the fall in the relative wage occurred before 1896, as did
almost all of the decline in the Irish emigration rate. But this simple
bivariate correspondence should not be pushed too far, because Irish
wages relative to American fell after 1896 yet the emigration rate did not
accelerate. More formal tests of the relative wage hypothesis can be
offered by time series regression analysis.

11
6 Grada, "Irish Emigration to the United States," p. 100.
12
For details see Williamson, "The Evolution of Global Labor Markets."
580 Hatton and Williamson

United States/Ireland

1
v
2P
3 2.0
2

1.5
Great Britain/Ireland

1 0 111111111111111111111111111 1111 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1850 1660 1870 1880 1890 1900 1910
Year
FIGURE 2

REAL WAGE RATIOS: UNITED STATES AND GREAT BRITAIN TO IRELAND

IRISH EMIGRATION: A TIME SERIES MODEL

There is a long research tradition for explaining emigration from the

Old World to the New. A number of studies have examined the
movements of total emigration from a given country, total immigration
to a given country, bilateral flows, and multicountry flows.13 They have
typically modeled migration flows as depending on wage rates (or per
capita income) and unemployment rates (or economic activity indica-
tors) in sending and receiving countries. The debate typically has
focused on three questions: First, were wage rates or levels of economic
activity the more significant variables? Second, did conditions in the
receiving countries have a larger or more significant influence than those
in the sending country?14 Third, can the large and significant coefficients
13
See, for example, Kelley, "International Migration and Economic Growth"; Gallaway and
Vedder, "Emigration from the United Kingdom"; Pope, "Empire Migration to Canada, Australia
and New Zealand" and "The Push-Pull Model of Australian Migration"; Richardson, "British
Emigration and Overseas Investment"; Wilkinson, "European Migration to the United States";
and Williamson, "Migration to the New World."
14
A number of writers have discussed this issue in terms of "push" and "pull" forces, typically
making inferences simply from the size of the estimated coeflBcients on sending and receiving
 Irish Emigration, 1850-1913 581
4.0

3.5

3.0

I
I 2.5

"3
2.0

1.5

Australia/Ireland
1.0 11111111111111111111111111111111111111111
1B50 1860 1870 1880 1890 1900 1910
Year
FIGURE 3

REAL WAGE RATIOS: AUSTRALIA AND CANADA TO IRELAND

typically found on the lagged dependent variable be interpreted as

reflecting what is called "chain migration"?
We follow established tradition by exploring the proximate causes of
emigration using relative real wages and employment prospects. How-
ever, our approach departs in some significant ways from the previous
literature. Our model can be expressed as follows:
MIP, = ao + a2Mog(ERh)t + a3Mog(WflWh),
a5log(ERh),-1 + a^og(Wf/Wh)t- ,+ ai(MSTlP),

+ as(M/P),-l
where MIP is the emigration rate, ERf is the foreign employment rate,
ERh is the home employment rate, Wf and Wh are respectively the
foreign and home real wage rates, and MST is the stock of previous
country right-hand-side variables. A more useful approach is to ask what share of any change in
migration was due to changing conditions at home and what share to changing conditions abroad,
either in terms of the proximate determinants of emigration or, better still, the underlying
demographic or labor market "fundamentals." See Williamson, "Migration to the New World."
582 Hatton and Williamson

emigrants living abroad. This model can be derived from utility-

maximizing microfoundations, and a full derivation can be found in
Timothy Hatton's study of United Kingdom emigration.15
There are a number of points worth noting about the model. First, the
dependent variable is the emigration rate—that is, the number of
emigrants relative to the population at risk rather than just the total
number of emigrants. Second, the relative wage term appears as the log
of the wage ratio, which constrains the home and overseas wage
coefficients to take equal and opposite signs. As a number of observers
have pointed out, emigration must surely have been driven by some
comparison, however approximate, between potential incomes at home
and abroad. In contrast to the expected income approach pioneered by
Michael Todaro and others, the overseas and home employment rates
are entered separately.16 This allows for differences in uncertainty about
gaining employment abroad. If emigrants were risk averse, the terms
reflecting the probability of employment in the overseas destination are
likely to take on larger coefficients than those for the wage ratio. The
coefficients on the home employment rate would be expected to take a
smaller coefficient than that on foreign employment if prospects at home
were more certain. Third, the relative wage and employment rates enter
as both changes and lagged levels. Although the emigration decision is
based on an assessment of the distant future, the timing of emigration
should depend on short-run changes in economic conditions. If we
characterize the long run by setting changes in wage and employment
rates equal to zero and {MlP)t = (MIP),_l, we emerge with the
following expression (dropping the time subscripts):
an dd as Q-c,
MIP = + \og{ERf) + - log(£R*) + log(Wf/Wh)
1 — ag 1 — ag 1 - ag 1 - a%
+ -^—(MST/P)
1 -a8
Thus, the long-run effect of a change in the relative wage on migration
is measured by the coefficient aj{\ - a 8 ). 17
The emigrant stock variable has been introduced to capture the
much-discussed phenomenon of chain migration. It has been argued
that past migration leads to future migration through the "friends and
relatives" effect. These contacts are thought to reduce the uncertainty
associated with emigration, to lower emigration costs by the remittance
15
Hatton, "A Model of UK Emigration."
16
Todaro, "A Model of Labor Migration."
17
The immigrant stock is itself an endogenous variable in the long run, so the true long-run
equation should account for this, too. In principle there ought to be an equilibrium stock/flow
relationship, depending on vital rates, linking the emigration rate and the population stock at home
and abroad.
 Irish Emigration, 1850-1913 583

of tickets or fares, and to reduce the costs of a job search on arrival.18

According to Fitzpatrick, chain migration "became the main agency
trans-planting the young adults of each generation." 19 These effects are
sometimes thought to be captured by the lagged dependent variable, but
the total stock of emigrants may be a more appropriate indicator of
them. Even so, the lagged dependent variable may be important to the
extent that it captures the formation of expectations through a simple
adaptive process. 20
The model is estimated on annual observations for 1877 to 1913, the
period for which the data are thought to be reasonably reliable (the
series for emigration are from Ferenczi and Willcox).21 The overseas
wage and employment rates are (geometric) weighted averages of the
United States, Great Britain, Canada, and Australia, with the weights
reflecting emigration shares.22 Internationally comparable real wage
rates for these countries and Ireland are from Jeffrey Williamson.23 The
U.S. employment rate is taken from J. R. Vernon; that for Britain, from
Charles Feinstein; and that for Australia, from Noel Butlin.24 The
employment rate for Canada was constructed on the basis of move-
ments in real GNP per capita.25 No series are available for Irish
unemployment, so domestic employment fluctuations are proxied by
deviations of the log of agricultural output from trend.26 We constructed
the stock of previous emigrants by using the emigration series to
interpolate between census benchmarks for the Irish-born living in the
four countries already mentioned and in New Zealand.
The regressions in Table 2 are for total emigration (the same series
graphed in Figure 1) and for various subaggregates separately. The
terms for the change in the wage ratio and the change in farm output
never took on significant coefficients in any of the regressions; there-
fore, we omitted them in the reported regressions. In initial experiments
the proportion of the Irish population between 15 and 34—the major
emigration age group—was also included, but this variable (interpolated
between census benchmarks) was never significant in any of the
18
Schrier, Ireland and the American Emigration, chap. 2.
19
Fitzpatrick, Irish Emigration, 1801-1921, p . 2 1 .
20
Hatton, " A Model of U K Emigration," p . 9.
21
Ferenczi and Willcox, International Migrations, p . 731.
22
The weights a r e United States, .7; Great Britain, .2; Canada and Australia, each .05. In the
estimates presented hereafter for intercontinental emigration, the weights are United States, .875;
Canada and Australia, each .0625.
23
Williamson, "The Evolution of Global Labor Markets."
24
Vernon, "Unemployment Rates in Post-Bellum America"; Feinstein, National Income,
Expenditure and Output; and Butlin, "An Index of Engineering Unemployment."
25
The variable was constructed by regressing the log of the employment rate on the deviation of
real per capita G N P from its logarithmic trend for 1916 t o 1926 and then extrapolating backward.
The employment rate w a s taken from Galenson and Zellner, "International Comparison of
Unemployment R a t e s , " p . 455. T h e real G N P is from Urquhart, " N e w Estimates of Gross
National P r o d u c t , " pp. 30-31.
26
Obtained from Drescher, " T h e Development of Agricultural P r o d u c t i o n . "
584 Hatton and Williamson
TABLE 2
DETERMINANTS OF TIME SERIES EMIGRATION RATES, 1877-1913
Dependent Variable
Emigration
to the Total
Total Adjusted Intercontinental United Total Male Female
Emigration Total Emigration States Emigration Emigration
Constant -11.61 -17.46 -8.27 -8.58 -11.37 -11.76
(2.56) (2.94) (2.01) (2.11) (2.38) (2.63)
Change in foreign 108.95 119.49 78.17 64.80 132.90 85.51
employment (4.52) (4.84) (3.13) (3.18) (5.12) (3.65)
rate
Foreign 64.57 74.68 50.33 35.75 76.60 54.16
employment (2.63) (2.90) (2.15) (1.84) (2.78) (2.35)
rate (/ - 1)
Deviation of log -17.24 -14.57 -22.69 -22.52 -14.87 -19.61
farm output (2.65) (2.20) (3.24) (3.58) (2.12) (3.11)
( « - 1)
Relative wage 13.16 17.94 8.39 7.20 16.41 9.98
(t- 1) (3.88) (4.44) (2.45) (2.42) (4.17) (3.20)
Migrant stock 22.87 26.54 22.46 25.97 20.31 25.10
(2.29) (2.57) (1-91) (1.98) (2.02) (2.43)
Lagged dependent 0.44 0.37 0.52 0.57 0.45 0.45
variable (3.46) (2.75) (4.00) (4.54) (3.51) (3.38)
R2 0.86 0.88 0.81 0.83 0.87 0.85
Residual sum of 80.05 83.45 96.26 76.40 92.73 75.56
squares
Durbin-Watson 1.90 1.84 1.88 2.00 1.98 1.83
LM(1) 0.15 0.31 0.22 0.00 0.00 0.40
Notes: The r-statistics are given in parentheses. LM(1) is the Lagrange Multiplier test for serial
correlation, obtained from the regression of residuals on lagged residuals and right-hand-side
variables. Foreign wage and employment rate variables and the migrant stock term are defined
differently across equations to be consistent with the destinations reflected in the dependent
variable.

regressions. Neither did a time trend added to the regressions take on a

significant coefficient. It appears, therefore, that Irish emigration can
largely be explained by the variables included in Table 2, which account
for four-fifths of the variance in the emigration rate.
The coefficients on the first regression reported in Table 2 imply very
powerful effects arising from the overseas employment rate. In the short
run, a 10 percent change in the employment rate (for example, a change
in the unemployment rate from 10 to 1 percent) would raise the
emigration rate by more than 11 per 1,000. If sustained in the long run,
it would raise the emigration rate by a similar magnitude. A rise of 10
percent in the foreign-to-domestic wage ratio, if sustained, would lead
ultimately to an increase in the emigration rate of 2.35 per 1,000.
Although this elasticity is much smaller than that for the employment
rate, the actual change in the wage ratio over the period was much larger
than the change in the overseas employment rate. Similarly, a 10
 Irish Emigration, 1850-1913 585

percent deviation of farm output from trend would tend to lower

emigration by about 3 per 1,000 in the long run. The migrant stock term,
though only bordering on significance, indicates in the long run that for
every 1,000 previous migrants an additional 41 were attracted overseas
each year.
We noted earlier that the statistics underestimate the flow of migrants
to Britain. In the second column of Table 2 the emigration rate has been
adjusted by doubling the recorded flow to Britain. This serves to slightly
raise the size and significance of the estimated coefficients on the
relative wage and overseas employment rates, though the long-run
effects are little changed. The third column in the table is for intercon-
tinental emigration alone (excluding Britain). The coefficients diminish
in size slightly, but the pattern is similar. The same applies to the fourth
regression, which is for emigration to the United States alone. In
particular, it seems that much of the short-run variation in emigration
can be attributed to changing employment conditions in the United
States. 27
The final two regressions in Table 2 indicate that the model fits well
for both male and female emigrants. The result for females is particu-
larly striking, given that the wage rates are for male occupations. It
suggests that our wage and employment variables are good proxies for
conditions relevant to female migration decisions. However, the results
do suggest that females were slightly less responsive to variations in
overseas employment rates and the relative wage than were males, but
slightly more responsive to variations in domestic activity.
To further confirm the importance of the role of relative wages rates
in the regressions of Table 2, we performed two other tests. First, for
each of the Table 2 regressions, we tested the restriction that the logs of
the foreign and home wage rates were equal and had opposite signs.
This restriction was easily accepted at the 5 percent level on a
chi-squared test for each of the equations. This confirms the importance
of the relative wage rather than the Irish wage alone. Second, we
experimented with an index of Irish agricultural wage rates, though we
do not have the appropriate price deflator. The results gave generally
larger though somewhat less significant coefficients. Thus, in the
equation for total emigration, the coefficient was 20.5 with a t-value of
2.3. This again confirms the importance of the relative wage.
What do these results imply for the long-run fall in the emigration
rate? Using the coefficients from the first column of Table 2, a 10 percent
27
We also experimented with separate regressions for the comparatively small flows to Canada
and Australia. Although these provided some support for the model, the equations did not perform
as well as for the aggregates or for the United States. We suspect that this is partly because the
flows to these countries were influenced by conditions in the United States and Britain. However,
because of collinearity between the variables, we were not able to model successfully the impact
of competing destinations.
586 Hatton and Williamson

fall in the wage ratio would cause the emigration rate to decline by 2.35
per 1,000. This is also consistent with the result for the United States
alone, which suggests that emigration there would decline by 1.7 per
1,000. Thus the 17 percent fall in the wage ratio between 1876/80 and
1909/13 accounts for a long-run fall in the emigration rate of 4 per
1,000—more than half of which would be accounted for by the 13
percent fall in the U.S./Irish wage ratio. Projecting this back to 1852/56,
the fall of 43 percent in the overall wage ratio must have lowered the
long-run emigration rate by as much as 10 per 1,000, accounting for
much of the secular decline in the emigration rate that we observe in the
data.
These results indicate that the fall in the ratio of overseas-to-Irish real
wages was central in accounting for the long-run decline in emigration
after 1850. Some of the decline after 1875 was due to the fall in the stock
of previous emigrants: from 59 percent of the Irish population in 1876/80
to 49 percent in 1909/13. To the extent that this was an indirect result of
the declining wage ratio, its impact on the decline in emigration rates
was even greater than these calculations imply.

WHO WERE THE IRISH MIGRANTS?

Scholars have long believed that we can gain a clearer picture of the
forces that drove so many people to emigrate by examining the
composition of the emigrants. But to draw inferences from this type of
evidence, the composition of the emigrants must be compared with the
population at risk. Even then, the evidence is more useful for examining
the factors that conditioned emigration in the country of origin than
those in the receiving country. Studies along these lines have been
undertaken for emigrants from England and Scotland by Charlotte
Erickson and for Ireland by Fitzpatrick and Patrick Blessing.28
Evidence on the composition of Irish emigrants again comes from the
emigration reports of the Registrar General for Ireland. Because the
accuracy of the returns before 1876—particularly by county of origin—
has been questioned, we concentrate on the period that followed.29
Although the continuing underenumeration of emigrants to Britain is
still present, it is unlikely to undermine seriously the findings from the
reported statistics. We concentrate chiefly on the census years from
28
Erickson, " W h o Were the English and Scottish Emigrants?"; Fitzpatrick, "Irish Emigra-
t i o n , " and Irish Emigration, 1801-1921; and Blessing, "Irish Emigration to the United S t a t e s . "
29
According to Fitzpatrick, " F o r the period between the great depression of the 1870s and the
First World War, greater reliance may be placed in the county distributions of recorded emigrants.
Enumeration was aided not only by the increasing popularity of emigration direct from Ireland to
N o r t h America, but also by the reorganisation of the count in 1876 which put an end to local failures
of enumeration due t o the unchecked laziness or incompetence of the responsible constables."
Fitzpatrick, "Irish Emigration in the Later Nineteenth Century," p . 28.
 Irish Emigration, 1850-1913 587

TABLE 3
EMIGRATION RATES PER 1,000 BY AGE GROUP AND REGION, IN CENSUS YEARS

1881 1891 1901 1911

Male Female Male Female Male Female Male Female
Age
0-4 7.6 7.5 4.8 4.7 3.8 3.9 2.2 2.1
5-9 6.2 6.2 3.5 3.6 2.6 2.7 1.8 1.5
10-14 5.0 5.9 2.9 3.0 1.7 2.0 1.6 1.6
15-19 21.0 35.1 17.8 31.8 6.8 16.6 9.7 18.9
20-24 67.1 55.1 59.3 53.3 34.9 41.2 37.8 29.0
25-34 31.6 16.9 23.1 13.3 16.6 12.7 16.8 8.5
35-^4 8.9 6.2 6.5 4.8 5.4 5.3 3.2 2.0
45-54 5.1 4.5 3.5 3.1 2.3 2.2 1.3 1.1
55+ 1.5 1.4 0.8 0.8 0.8 0.9 0.3 0.4
Total 15.9 14.5 13.0 12.4 8.3 9.5 7.8 6.4
United United United United
States Other States Other States Other States Other
Region
Leinster 10.9 1.8 6.9 0.8 2.4 0.6 2.6 1.1
Minister 12.7 3.7 18.3 2.9 12.4 2.8 6.8 0.6
Ulster 8.8 5.0 6.7 1.5 3.1 2.4 3.5 4.1
Connaught 18.7 1.3 16.3 0.9 16.7 0.4 10.5 0.4
Sources: Calculated from Emigration Statistics ofIreland and Census of Ireland. See the Appendix
for further details.

1881 to 1911, so that the emigrants can be compared with the source
population.

Gender and Age

Unusually, in comparison with male-dominated emigration from
other European countries, Irish emigration was composed almost
equally of males and females. Females accounted for 48 percent of all
recorded emigrants over the period as a whole. As the first panel of
Table 3 shows, the emigration of both sexes was highly concentrated in
the age groups from 15 to 34, and especially in the age group of 20 to 24.
The 15 to 34 age group comprised over 80 percent of the emigrants. In
1881 alone over 60 per 1,000 of those in the 20 to 24 age group emigrated.
The main difference between male and female emigrants is that emigra-
tion rates were higher for women in the 15 to 19 age group and lower in
the 25 to 34 age group.
The total number per 1,000 emigrating in the 15 to 34 age group fell
from 35.9 in 1881 to 25.2 in 1911. Indeed, the decline we have observed
in the aggregate emigration rate was common to all age groups; hence,
the age structure of the emigrants varied little over time. It has been
calculated that, for the cohort aged about 15 in 1876, the probability of
588 Hatton and Williamson

remaining in Ireland until the age of 55 was only about .53 for men and
.58 for women. 30
Most of the emigrants were single: the proportion of all emigrants
aged 15 and over who were married or widowed was only 10 to 15
percent. Children under 15 typically emigrated as part of a family group.
The lower emigration rates for those aged 10 to 14 compared with the 0
to 4 age group reflects the lower emigration rates of parents of older
ages.
Region of Origin
Where did these emigrants come from? Emigration rates varied
greatly across different regions. They were higher in the rural counties
to the south and west. As the second panel in Table 3 shows, at the
provincial level emigration rates were much higher from Connaught and
Munster than from Ulster and Leinster; but there was also considerable
variation within provinces. What stands out most is the low emigration
rates from County Dublin and to a lesser extent from the most urbanized
parts of Ulster, such as Counties Antrim and Down. The overall
variation in county emigration rates increased and then decreased over
the period from 1881, with no strong evidence of the diffusion effect
identified by J. D. Gould for Italy and Hungary.31 The coefficient of
variation of county emigration rates was .31 in 1881, .44 in 1891, .71 in
1901, and .41 in 1911.32 Furthermore, there was considerable persis-
tence in county emigration rates: the correlation coefficient between the
rates in 1881 and 1911 was .42.
Table 3 also shows the numbers per thousand emigrating to the
United States and to all other destinations combined. Although the
latter figure is downward biased, the differences across provinces are
clear. Emigrants from Leinster, Munster, and Connaught showed a
much greater preference for the United States than for other destina-
tions. Compared with those leaving other provinces, those from Ulster
showed a greater preference for Great Britain, Canada, Australia, and
New Zealand.
Occupation
The reports on emigration also give a breakdown of the occupational
structure of the emigrants, though these are somewhat harder to
compare with census data because of differences in classification. It is
clear, however, that among emigrants the unskilled were overrepre-
sented relative to the occupied population. Of emigrating males who
30
Fitzpatrick, "Irish Emigration."
31
Gould, " E u r o p e a n Inter-Continental Emigration: The Role of Diffusion and F e e d b a c k . "
32
This result does not depend on o u r selection of these specific years. If we take average
emigration rates for intercensal periods, the coefficients of variation are .25 for the 1870s, .31 for
the 1880s, .60 for the 1890s, and .46 for the 1900s.
 Irish Emigration, 1850-1913 589

reported an occupation, the proportion of laborers was 80 percent in

1881, falling to 65 percent in 1911. By contrast, among the occupied
population recorded at the census, 22 percent in 1881 and 21 percent in
1911 were laborers. Similarly, of the female emigrants who reported an
occupation, the vast majority were servants: 84 percent in 1881 and 85
percent in 1911. This compares with 33 percent of the occupied female
population at the 1881 census and 34 percent in 1911. Thus, the
emigration rates for these occupations were very high. In 1911 the
emigration rates were 35 per 1,000 for male laborers and 26 per 1,000 for
female servants.
What inferences can be drawn from these statistics? Those who
emigrated were bound to be more mobile than the population at large,
because the costs of migrating were lower for the young and single.
They had little investment in skills that were specific to firms, to
industries, or to Ireland as a whole. It seems likely that the present value
of emigration was much greater for these young, unskilled workers than
for those who were more established, more skilled, and older.
It has sometimes been argued that diminishing opportunities in
agriculture were a major reason for emigration. This would be consis-
tent with the observed high emigration rates from counties in the rural
west and south. The decline of the potato culture, underemployment,
low wages, and limited opportunities to move up the agricultural ladder
to become a small farmer are all seen as contributing to the flow of
emigration from rural areas. The decline of the small farm probably
made opportunities for women even more meager than for men.
Fitzpatrick found a strong correlation across counties between the
emigration rate and the population resident on agricultural holdings, and
also between the emigration rate and the decline in the proportion of
agricultural laborers in the population.33

Cultural and Economic Factors

There has been considerable debate about the link between the
system of inheritance in rural households and emigration. It has been
argued that the "stem family" system prevailed in Ireland, though its
extent has been debated.34 After the Famine, the division of farms
became relatively rare; hence, one child inherited the family farm,
leaving the others to seek their fortunes elsewhere. The inheritor was
generally a son, though not necessarily the eldest.35 In a recent study of
turn-of-the-century household data, Timothy Guinnane has shown that
33
Fitzpatrick, "Irish Emigration," p. 138.
34
Gibbon and Curtin, "The Stem Family in Ireland"; Fitzpatrick, "Irish Farming Families";
and Guinnane, "Intergenerational Transfers."
35
C Gr&da, "Primogeniture and Ultimogeniture"; and Kennedy, "Farm Succession in Modern
Ireland."
590 Hatton and Williamson

the incidence of departure among sons and daughters depended on

economic and demographic characteristics of the family, including the
age of the household head and the number of siblings.36 In general, we
would expect large families to be associated with high rates of emigra-
tion. The prospect of emigration was surely preferable to the low wages
and underemployment of the farm laborer.
Some writers have argued that extreme poverty both encouraged
emigration and hindered it. Though the poor would have had most to
gain from the move, they would have been least able to finance it. S. H.
Cousens, for instance, found that emigration rates were more equal
across counties in the 1860s and 1870s than they were later in the period.
He argued that "the very poverty of these western areas partly
explained the lack of mass emigration."37 In later decades, rates of
emigration were higher from the west than from other areas. This
change might have been due to the underenumeration of emigrants from
the west in the earlier decades, or to the relaxation of the poverty
constraint through remittances of funds or tickets from friends and
relatives who had emigrated previously. It is hard to believe that
poverty seriously hindered emigration from the west after the 1870s,
given that such a large proportion of the emigrants chose the more
expensive passage to America rather than the more modest investment
of emigration to Britain.38
Other factors may also have acted to constrain emigration from the
more backward areas. Illiteracy and an inability to speak English may
have limited knowledge of opportunities abroad as well as reducing the
potential gains from emigration. This could have affected a significant
number of potential emigrants: among the population over the age of
five in 1881, one-third could neither read nor write, and 3.4 percent
spoke only Irish. By 1911 these percentages had fallen to 8 percent (aged
nine and over) who were illiterate and 1.5 percent who spoke only Irish.
If these characteristics served to retard emigration, then their effects
must have become weaker over time.
Religious and Political Factors
A final set of characteristics sometimes associated with emigration
relate to religion and republicanism. Despite some suggestion to the
contrary, it is not clear that Catholics had a higher propensity to
emigrate when other factors are taken into account. But it has been
noted that Protestants and Catholics emigrated to different destinations.
Ulster Protestants tended to emigrate to Britain or Canada, whereas the
Catholics of the south and west disproportionately went to the United
States. It is unclear how far this reflects political preference. Fitzpatrick
36
Guinnane, " A g e at Leaving H o m e . "
37
Cousens, " T h e Regional Variations in Population Changes," p . 311.
38
Fitzpatrick, Irish Emigration, 1801-1921, p . 9; and 6 Grada, " S o m e A s p e c t s , " p. 70.
 Irish Emigration, 1850-1913 591

has argued that this distribution arose from regional differences in

emigration established after the Famine and sustained via chain migra-
tion.39
One problem with such observations is that in general they are
derived from simple bivariate comparisons or cross tabulations that do
not control for the full set of characteristics simultaneously. Hence it is
impossible to say, for example, whether the presence of industry, the
level of wages, or just the extent of urbanization acted as a deterrent to
emigration. Similarly, it is unclear how far poverty, illiteracy, and
differences in religion contributed to raising the emigration rate, as they
were correlated across regions and counties. To shed further light on
these questions, we must turn to cross-sectional regression analysis.

IRISH EMIGRATION: A CROSS-SECTIONAL MODEL

To examine the determinants of emigration, we extracted county data

for each of the single census years 1881, 1891, 1901, and 1911, yielding
a pooled cross section of 128 observations. This allowed us to use the
census for most of our explanatory variables. These years were gener-
ally close to the trend in emigration rates. 40 Using individual years
rather than, say, decade averages of emigration lets us avoid, as far as
possible, the problem that emigration itself influenced the demographic
and other variables that we obtain from the census.
The demographic variables include the proportion of the population in
the prime emigration age group of 15 to 34, the average family size, and
the number of persons per house. These are expected to be positively
correlated with emigration. The proportion of a county's population
who were born there is intended to test the hypothesis that those who
had moved once were more likely to move again. We expect the
proportion who were Catholic to raise the emigration rate, and the
proportion who were illiterate or who spoke only Irish to lower it.
The economic characteristics of the counties are represented by
several variables. The proportion living in towns of 2,000 or more
persons measures the degree of urbanization, whereas the proportion of
the labor force in industry and the proportion in agriculture reflect
differences in the occupational composition of the labor force. The
potential access to land in the form of a small farm is proxied by the
proportion of agricultural holdings of less than five acres. This variable
is interacted with the proportion of the male labor force in agriculture to
give it a greater weight in the more agricultural counties. Irish employ-
ment conditions are measured by the farm wage and, to capture the
39
Fitzpatrick, "Irish Emigration," p. 129.
40
The deviations from the linear trend between 1880 and 1913 for the aggregate emigration rate
were (in emigrants per 1,000 of population) - 1 . 1 5 in 1881, 0.07 in 1891, - 0 . 3 3 in 1901, and 1.31 in
1911.
592 Hatton and Williamson

incentive to emigrate, this is divided by a weighted average of the real

wage in destination countries.41 The extent of poverty is measured by
two variables: the proportion of families living in third- and fourth-class
housing and the proportion of the population receiving poor relief.42
Preliminary tests showed that a number of variables attained very low
significance levels and could be excluded. These were the proportion
born in the county of residence, the proportion of the labor force in
industry, the average number of persons per house, and the proportion
speaking Irish only. Regressions with the remaining variables for total
emigration, emigration to the United States, and intercontinental emi-
gration are presented in Table 4. These regressions perform well by
cross-sectional standards, explaining at least two-thirds of the variation
in the dependent variable in each case. We initially included year
dummies but found, on a chi-squared test for joint significance, that they
could be excluded in each case.
Turning to the regression for total emigration in the first column of
Table 4, the point estimate for the proportion of the population between
15 and 34 indicates, plausibly, that emigration would be 13.6 per 1,000
higher for this group than for the population as a whole—though
surprisingly it is not significant. The proportion of the population in
towns gives a positive sign, though it is small and also insignificant,
indicating that urbanization alone had little effect on the propensity to
emigrate. More important are the agricultural variables: the greater the
proportion of the labor force in agriculture, the greater was emigration;
the higher the proportion of smallholdings, the smaller was emigration,
given the share of the labor force in agriculture. These results strongly
suggest that lack of opportunities in agriculture (including opportunities
to obtain or inherit a small farm) was an important cause of emigration.
We constructed the relative wage term to facilitate comparison with
the time series results, and entered it in logarithmic form. The point
estimate is noticeably smaller than the short-run coefficient estimated
on time series in Table 2 and much lower than the long-run time series
coefficient. It suggests that a 10 percent rise in the relative wage would
raise the emigration rate by 0.64 per 1,000. However, the poverty
variables also have noticeably significant effects on emigration. An
increase in the number on poor relief by 1 percent of the population
41
The wage data were taken from Bowley, "The Statistics of Wages in the United Kingdom,"
p. 401, for 1880/81, and from Fitzpatrick, "The Disappearance of the Irish Agricultural Labourer,"
p. 90, for 1893 and 1911, with estimates for 1901 obtained by linear interpolation. For each census
year the wage ratios were adjusted so that the mean of the counties relative to the wage overseas
would be the same as that for builders' laborers used in the time series analysis described earlier.
42
This variable includes both indoor and outdoor relief. In preliminary tests we attempted to
adjust for differences in poor law policy between counties in the manner suggested by MacKinnon,
"English Poor L a w , " using the ratio of the number relieved indoors to the total relieved as an
indicator of policy stance. Because the results from the corrected relief variable were very similar
to the raw variable, we retained the latter.
 Irish Emigration, 1850-1913 593

TABLE 4
REGRESSION ESTIMATES FOR COUNTY EMIGRATION RATES, 1881-1911
Emigration Population Aged 15 to 34
tn thf>
IU UlC
Total United Intercontinental Total Male Female
Emigration States Emigration Emigration Emigration Emigration
Constant -45.92 -50.% -45.18 -113.78 -96.96 -129.84
(4.20) (5.20) (4.38) (5.19) (4.26) (5.39)
Proportion aged 15 13.60 21.82 16.44
to 34 (0.88) (1.58) (1.13)
Proportion in towns 3.02 3.17 2.03 3.88 -4.47 11.27
> 2,000 (0.50) (0.60) (0.36) (0.26) (0.28) (0.68)
Proportion of male 16.39 15.90 15.46 41.60 34.28 48.03
labor force in (2.38) (2.56) (2.37) (2.46) (1.95) (2.59)
agriculture
Proportion of -41.98 -42.42 -46.50 -127.48 -123.63 -130.65
holdings < 5 (3.26) (3.68) (3.84) (3.91) (3.65) (3.65)
acres • proportion
of labor force in
agriculture
Log of foreign 6.43 6.49 7.54 18.86 21.23 16.26
wage relative to (2.12) (2.33) (2.61) (2.48) (2.69) (1.95)
Irish wage
Percentage of 1.96 1.56 1.88 4.14 5.78 2.51
population on (4.40) (3.95) (4.53) (3.67) (4.95) (2.03)
poor relief
Proportion of third- 7.69 8.88 9.17 16.31 7.79 24.27
and fourth-class (1.45) (1.89) (1.86) (1.29) (0.60) (1.75)
houses
Average family size 7.93 7.15 6.83 21.22 20.55 22.07
(5.24) (5.30) (4.82) • (5.57) (5.20) (5.28)
Proportion Catholic -4.97 1.04 -1.43 -1.38 -14.07 10.59
(1.79) (0.42) (0.55) (0.20) (1.97) (1.40)
Proportion illiterate 1.53 -0.80 -1.13 -9.70 -3.30 -15.44
(0.19) (0.11) (0.15) (0.50) (0.16) (0.73)
R2 0.70 0.74 0.72 0.72 0.68 0.71
Residual sum of 1420.9 1188.6 1309.9 9646.6 10360.2 11590.0
squares
HETERO 0.73 2.99 1.77 1.09 1.53 1.49
Notes: The /-statistics are given in parentheses. HETERO is the test statistic for heteroscedasticity,
based on regressing squared residuals on squared fitted values. It is distributed as chi-squared with
one degree of freedom; the critical value at 5 percent equals 3.84.

would raise emigration by 1.96 per 1,000; a 1 percent increase in the

proportion of third- and fourth-class housing would raise emigration by
0.08 per 1,000.
Average family size has a large and highly significant coefficient. It
implies that a one-person reduction in average family size would lower
the emigration rate by 7.93 per 1,000. This would support those who
have argued that demographic forces continued to influence emigration
in the late nineteenth century. High birth rates led to large families,
which precipitated a large flow of emigration. Furthermore, it was
594 Hatton and Williamson

family size rather than household size that dominated in the regression
when both variables were included. Finally, the proportion of Catholics
has a negative sign and the proportion illiterate a positive sign, contrary
to expectation (though the coefficient is insignificant in the latter case).
Overall it appears that income, poverty, and demographic variables
were the key determinants of county emigration rates.
The second regression in Table 4 is for emigration to the United States
only, and the results are broadly similar to those for the aggregate.
There is some increase in the size of the coefficients on the proportion
of the population between 15 and 34 and on the proportion living in
towns, but these variables remain insignificant. The other main differ-
ence is the change in the coefficient on the proportion of Catholics,
which becomes positive but remains insignificant. Given the small size
and insignificance of this coefficient, it appears that Protestants and
Catholics were equally likely to emigrate to the United States, but that
Protestants were more likely than Catholics to emigrate to other
destinations.
Taking intercontinental emigration as a whole, the third regression
indicates some increase in the Catholic coefficient, though it is still
insignificant. Again, the other coefficients remain largely unchanged. As
we did for the time series analysis, we ran a regression for adjusted total
emigration, in which the recorded number emigrating to Britain was
doubled. The main difference in this result (not shown) was that the
coefficient on the Catholic variable became even more negative (-8.7)
and achieved a f-value of 2.8.
Because the migration data are available for different groups by age
and sex, we can explore the impact of the variables on the key
emigration group: those aged 15 to 34. The results are reported in the
last three columns of Table 4: for the most part the coefficients are
similar to those for all emigrants, but much bigger. Thus for the age
group of 15 to 34 alone, a 10 percent rise in the relative Irish wage would
raise the age-specific emigration rate by 1.9 per 1,000. Similarly, a rise
in the number on poor relief by 1 percent of the population would raise
the emigration rate by 4.1 per 1,000; a one-person increase in average
family size would raise it by 16 per 1,000.
When we treat males and females separately, only minor differences
emerge. The results for the relative wage again suggest that this variable
captures the wage incentives for both men and women. One point of
difference is that the share urban has a negative effect on male but a
positive effect on female emigration, though in neither case is the
coefficient significant. In addition, females were more likely to emigrate
the higher the proportion of the labor force in agriculture. A further
contrast is in the housing quality variable, which is larger and more
significant for females.
How much did each of these variables affect emigration over time?
 Irish Emigration, 1850-1913 595
TABLE 5
DECOMPOSITION OF CHANGES IN EMIGRATION RATES, 1881-1911
(effects on unweighted means of county emigration rates from first column in Table 4)
1881-1901 1881-1911
Proportion aged 15 to 34 0.26 -0.02
Proportion in towns > 2,000 0.08 0.11
Proportion of male labor force in agriculture -0.05 -0.29
Proportion of holdings < 5 acres • proportion of -0.68 -1.12
labor force in agriculture
Log of foreign wage relative to Irish wage -0.92 -0.82
Percentage of population on poor relief -2.08 -1.38
Proportion of third- and fourth-class houses -0.97 -1.63
Average family size -2.71 -3.63
Proportion Catholic 0.01 -0.02
Proportion illiterate -0.18 -0.20
Total change -7.24 -9.00
Actual change -8.44 -9.52

We can examine this issue by taking the means of the variables across
counties for pairs of years and then, using our estimated coefficients,
decomposing the change in the (unweighted) mean of county emigration
rates between two periods. The unweighted means for each census year
are given in the Appendix. We focus on the two periods from 1881 to
1901 and 1881 to 1911 because some of the variables, specifically the
relative wage and the proportion on poor relief, changed their trend in
the last decade. We use the first column of Table 4 to decompose
changes in the total emigration rate.
This decomposition is reported in Table 5. The results indicate, once
again, that the key variables reducing the emigration rate over time were
the demographic variables and those reflecting living standards. The
agricultural variables—the proportion of the labor force in agriculture
and this same variable interacted with the proportion of holdings of less
than 5 acres—together contributed modestly to the fall in the average
emigration rate, reducing it by 1.4 per 1,000 over the whole period. The
decline in the number of persons per family had strong negative effects
on the emigration rate, reducing it by 2.7 per 1,000 in 1881/1901 and 3.6
per 1,000 in 1881/1911. If the relative wage, the proportion of the
population on poor relief, and housing quality together are taken to
reflect living standards, these contributed even more to the decline in
the emigration rate. They account for a fall of 4.0 per 1,000 in 1881/1901
and 3.8 per 1,000 in 1881/1911.
These cross-sectional findings may seem at first sight to be at variance
with the time series results, which laid stress on the importance of
relative wage rates in driving Irish emigration. The cross-sectional
results suggest that wage rates alone had a relatively small impact on the
fall in emigration, especially over the whole period from 1881 to 1911.
But wage rates can be viewed as a summary statistic representing
596 Hatton and Williamson
TABLE 6
RESTRICTED REGRESSIONS FOR COUNTY EMIGRATION RATES, 1881-1911

Total Emigration to the Intercontinental

Emigration United States Emigration

Constant -54.42 -57.85 -55.83

(10.07) (11.34) (10.52)
Log of foreign wage 11.91 12.74 13.14
relative to Irish wage (4.15) (4.61) (4.62)
Average family size 12.17 12.21 11.92
(10.58) (11.59) (10.82)
R2 0.57 0.59 0.57
Residual sum of squares 2145.9 1866.1 2030.4
HETERO 0.66 2.09 1.62

Notes: See Table 4.

various aspects of living standards. Thus, the different relative wage

effects may be more apparent than real. We can easily test this assertion
by estimating a much-reduced version of the equation in Table 4, one
that includes only the family size variable and the log relative wage.
The results, presented in Table 6, suggest that the wage does indeed
capture a range of variables representing living standards. The coeffi-
cient of 11.9 on the wage in the first column is nearly twice as large as
it was in Table 4. This indicates that a 10 percent rise in the relative
wage would reduce emigration by 1.2 per 1,000, about the same as the
short-run coefficient in the time series equation and about half the
long-run coefficient. If the family size variable is excluded, the wage
coefficient rises still further, though it is still a bit smaller than the
long-run time series estimates.43 Hence, it appears that the time series
and cross section results can be largely reconciled.

CONCLUSION

We have argued here that the secular decline in the Irish emigration
rate from its immediate post-Famine level owes a good deal to the
gradual rise in Irish wage rates and living standards up to 1913. Time
series analysis suggests that much of the decrease in emigration to the
United States can be accounted for by the convergence of the Irish real
wage with the American. The cross-sectional evidence argues, how-
ever, that living standards should be construed more widely than simply
as real wage levels. Just as they had during and immediately after the
Famine, the Irish in the late nineteenth century were still emigrating to
43
The result for total emigration is

MIP= 0.42 +19.73 \og(WfIWh) fl2 = 0.18

(0.20) (5.19)
This should be compared with the long-run relative wage coefficient for total emigration, estimated
in Table 1 to be 23.5.
 Irish Emigration, 1850-1913 597

escape poverty, as reflected by the numbers on poor relief and the

proportion in low-quality housing.
The results of our cross-sectional analysis also shed light on a number
of hypotheses in the existing literature. First, our results strongly
support the view, emphasized in demographic studies, that large fami-
lies led to emigration. Though our family size variable is a crude
indicator, the results are consistent with the hypothesis that one or two
children either inherited or married an inheritor, but the "surplus"
siblings typically faced the choice of becoming poor landless laborers or
emigrants. Second, and consistent with the first point, we find that
agricultural variables were important in determining cross-county emi-
gration rates. In particular, the greater the access to smallholdings, as
reflected in their proportion of the total of all land holdings, the lower
the emigration rate. Third, we find little evidence, at least in the period
after 1880, that poverty inhibited emigration. Both the numbers on poor
relief and the proportion of low-quality housing had positive effects on
emigration. We suspect that most prospective emigrants could obtain
help from relatives abroad, thus easing the poverty constraint. Finally,
there is little evidence that personal characteristics such as religion or
illiteracy were important determinants of emigration. But there is some
evidence that Catholics were less likely than non-Catholics to emigrate,
especially to destinations other than the United States.
Our findings, which stress the role of Irish living standards as a cause
of declining Irish emigration, also raise an important issue for future
research. To what extent did the decline in population and labor force,
driven by continuing emigration, lead to rising Irish living standards?
Almost all of the population decline affected agriculture, and, as Arnold
Schrier noted 30 years ago, many contemporaries associated it with the
rise in wages.44 More recently Kevin O'Rourke has argued that between
1856 and 1876, rural depopulation and changes in the structure of
agriculture, particularly the reduction in tillage, can be explained by
increasing labor scarcity due to external labor market conditions.45 But
as the wage gap between Ireland and the countries of immigration
declined, the Irish emigration rate decreased, as did the rate of decline
in population and the rates of increase in real wages and living
standards. If further research provides support for the link between
population and the real wage, then it would suggest that emigration was
an important factor driving the long-run convergence of wage levels in
an increasingly integrated international labor market.

44
Schrier, Ireland and the American Emigration, pp. 73-75.
43
O'Rourke, "Rural Depopulation," p. 428.
598 Hatton and Williamson

Appendix
M E A N S O F VARIABLES USED IN CROSS-SECTIONAL REGRESSION

Variable 1881 1891 1901 1911 Pooled

Emigration rate 16.55 12.63 8.11 7.03 11.08
Proportion aged 15 to 34 0.32 0.34 0.35 0.32 0.33
Proportion in towns >2,000 0.18 0.19 0.20 0.21 0.20
Proportion of male labor force in agriculture 0.66 0.65 0.66 0.65 0.66
Proportion of holdings < 5 acres 0.17 0.17 0.19 0.21 0.19
Foreign wage relative to Irish wage 1.88 1.73 1.62 1.65 1.72
Percentage of population on poor relief 2.51 2.37 1.45 1.80 2.03
Proportion of third- and fourth-class houses 0.46 0.40 0.33 0.25 0.36
Average family size 5.10 4.93 4.76 4.64 4.86
Proportion Catholic 0.82 0.81 0.81 0.81 0.81
Proportion illiterate 0.25 0.18 0.13 0.11 0.17

Sources: The numbers of emigrants in the total by age and country of destination were taken from
Parliament, Emigration Statistics of Ireland: 1881 (C. 2828), 1891 (C. 6679), 1901 (Cd. 976), and
1911 (Cd. 6131). The total population, numbers aged 15 to 34, numbers in towns, proportion on
poor relief, average family size, proportion Catholic, and proportion illiterate are from Parliament,
Census of Ireland, Provincial Summary Tables: Leinster, 1881 (C. 3042), 1891 (C. 6575), 1901 (Cd.
847), 1911 (Cd. 6049); Munster, 1881 (C. 3418), 1891 (C. 6567), 1901 (Cd. 1058), 1911 (Cd. 6050);
Ulster, 1881 (C. 3204), 1891 (C. 6626), 1901 (Cd. 1123), 1911, (Cd. 6051); Connaught, 1881 (C.
3268), 1891 (C. 6685), 1901 (C. 1059), 1911 (Cd. 6052). The proportion of holdings less than five
acres and the proportion of third- and fourth-class housing are from Parliament, Census of Ireland,
General Report: 1881 (C. 3365), 1891 (C. 6780), 1901 (Cd. 1190), 1911 (Cd. 6663). The proportion
of the male labor force in agriculture is from Fitzpatrick, "The Disappearance of the Agricultural
Labourer." For the derivation of the wage ratio, see the text.

REFERENCES
Blessing, Patrick L., "Irish Emigration to the United States, 1800-1920: An Overview,"
in P. J. Drudy, ed., The Irish in America: Emigration Assimilation and Impact
(Cambridge, 1985), pp. 11-37.
Bowley, A. L., "The Statistics of Wages in the United Kingdom During the Last
Hundred Years: Part 3, Agricultural Wages," Journal of the Royal Statistical
Society, 62 (June 1899), pp. 395-404.
Butlin, Noel G., "An Index of Engineering Unemployment, 1852-1943," Economic
Record, 22 (Dec. 1946), pp. 241-60.
Cousens, S. H., "The Regional Variations in Population Changes in Ireland 1861—
1881," Economic History Review, 17 (Aug. 1964), pp. 301-21.
Drescher, Leo, "The Development of Agricultural Production in Great Britain and
Ireland from the Early Nineteenth Century," The Manchester School, 23 (May
1955), pp. 153-75.
Erickson, Charlotte J., "Who Were the English and Scottish Emigrants in the 1880s?"
in D. V. Glass and Roger Revelle, eds., Population and Social Change (London,
1972), pp. 347-82.
Feinstein, Charles H., National Income, Expenditure and Output of the United
Kingdom 1855-1965 (Cambridge, 1972).
Ferenczi, Imre, and Walter F. Willcox, International Migrations: Vol. 1, Statistics
(New York, 1929).
Fitzpatrick, David, "Irish Emigration in the Later Nineteenth Century," Irish Histor-
ical Studies, 22 (Sept. 1980), pp. 126-43.
 Irish Emigration, 1850-1913 599
Fitzpatrick, David, "The Disappearance of the Irish Agricultural Labourer, 1841-
1912," Irish Economic and Social History, 7 (1980), pp. 66-92.
Fitzpatrick, David, "Irish Farming Families Before the First World War," Comparative
Studies in Society and History, 25 (July 1983), pp. 339-74.
Fitzpatrick, David, Irish Emigration, 1801-1921 (Dublin, 1984).
Galenson, Walter, and Arnold Zellner, "International Comparison of Unemployment
Rates," in NBER, The Measurement and Behavior of Unemployment (Princeton,
1957).
Gallaway, Lowell E., and Richard K. Vedder, "Emigration from the United Kingdom
to the United States: 1860-1913," this JOURNAL, 31 (Dec. 1971), pp. 885-97.
Gibbon, P., and C. Curtin, "The Stem Family in Ireland," Comparative Studies in
Society and History, 20 (July 1978), pp. 429-53.
Gould, J. D., "European Inter-Continental Emigration: The Role of Diffusion and
Feedback," Journal of European Economic History, 8 (Fall 1980), pp. 267-315.
Guinnane, Timothy W., "Intergenerational Transfers, Emigration, and the Rural Irish
Household System," Explorations in Economic History, 29 (Oct. 1992), pp. 456-76.
Guinnane, Timothy W., "Age at Leaving Home in Rural Ireland, 1901-1911," this
JOURNAL, 52 (Sept. 1992), pp. 651-74.
Hatton, Timothy J., "A Model of UK Emigration, 1870-1913" (Discussion Paper No.
771, Centre for Economic Policy Research, London, Mar. 1993).
Kelley, Allen C , "International Migration and Economic Growth, Australia: 1865-
1935," this JOURNAL, 25 (Sept. 1965), pp. 333-54.
Kennedy, Liam, "Farm Succession in Modern Ireland: Elements of a Theory of
Inheritance," Economic History Review, 44 (Aug. 1991), pp. 477-99.
Kennedy, Robert E., The Irish: Emigration, Marriage and Fertility (Berkeley, 1973).
MacKinnon, Mary, "English Poor Law Policy and the Crusade Against Outrelief," this
JOURNAL, 47 (Sept. 1987), pp. 603-26.
Miller, Kerby A., Emigrants and Exiles: Ireland and the Irish Exodus to North America
(New York, 1985).
Miller, Kerby A., "Emigration as Exile: Cultural Hegemony in Post-Famine Ireland,"
in Rudolph J. Vecoli and Suzanne M. Sinke, eds., A Century of European
Migrations, 1830-1930 (Urbana, IL, 1991), pp. 339-63.
Mokyr, Joel, Why Ireland Starved (London, 1983).
6 Grlda, Cormac, "A Note on Nineteenth Century Emigration Statistics," Population
Studies, 29 (Mar. 1975), pp. 143-49.
6 Grada, Cormac, "Some Aspects of Nineteenth Century Irish Emigration," in L. M.
Cullen and T. C. Smout, eds., Comparative Aspects of Scottish and Irish Social
History, 1600-1900 (Edinburgh, 1978), pp. 64-73.
6 Grdda, Cormac, "Irish Emigration to the United States in the Nineteenth Century,"
in David N. Doyle and Owen D. Edwards, eds., America and Ireland 1776-1976
(Westport, CT, 1980), pp. 93-104.
6 Gr&da, Cormac, "Primogeniture and Ultimogeniture in Rural Ireland," Journal of
Interdisciplinary History, 10 (Winter 1980), pp. 491-97.
O'Rourke, Kevin, "Rural Depopulation in a Small Open Economy: Ireland 1856-1876,"
Explorations in Economic History, 28 (Oct. 1991), pp. 409-32.
Pope, David, "Empire Migration to Canada, Australia and New Zealand, 1910-1929,"
Australian Economic Papers, 1 (Dec. 1968), pp. 167-88.
Pope, David, "The Push-Pull Model of Australian Migration," Australian Economic
History Review, 14 (Sept. 1976), pp. 144-52.
Richardson, Harry W., "British Emigration and Overseas Investment 1870-1914,"
Economic History Review, 25 (Feb. 1972), pp. 99-113.
Schrier, Arnold, Ireland and the American Emigration, 1850-1900 (Minneapolis, MN,
1958).
600 Hatton and Williamson
Todaro, Michael P., "A Model of Labor Migration and Urban Unemployment in Less
Developed Countries," American Economic Review, 59 (Mar. 1969), pp. 138-48.
Urquhart, M. C , "New Estimates of Gross National Product, Canada, 1870-1926:
Some Implications for Canadian Development," in Stanley L. Engerman and
Robert E. Gallman, eds., Long-Term Factors in American Economic Growth
(Chicago, 1986), pp. 9-94.
Vernon, J. R., "Unemployment Rates in Post-Bellum America, 1869-1899" (Unpub-
lished paper, University of Florida, 1991).
Wilkinson, Maurice, "European Migration to the United States: An Econometric
Analysis of Labor Supply and Demand," Review of Economics and Statistics, 52
(Aug. 1970), pp. 272-79.
Williamson, Jeffrey G., "Migration to the New World: Long Term Influences and
Impact," Explorations in Economic History, 11 (Summer 1974), pp. 357-89.
Williamson, Jeffrey G., "The Evolution of Global Labor Markets in the First and
Second World since 1830: Background Evidence and Hypotheses" (Unpublished
paper, Harvard University, 1992).